Repository navigation
Conversation
This comment has been minimized.
This comment has been minimized.
6 tasks done
6 tasks done
5 of 8 tasks
neilsh
added a commit
to neilsh/stunt-reel
that referenced
this pull request
Nov 15, 2026
rehype-katex 7 still declares katex ^0.16.0 while mermaid takes ^0.18.0, so two copies coexisted and the 0.16.47 pin left the renderer on a release without the advisory fixes. The override gives the tree one KaTeX, 0.18.11, inside mermaid's range; the direct pin keeps the stylesheet the same version. Renovate now allows below 0.19. Drop the override once remarkjs/remark-math#122 ships. The math spike (work/2026-11/576_research-site/math-spike.md on claude/spike-site-math-b43fe3) has the evidence. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
neilsh
added a commit
to neilsh/stunt-reel
that referenced
this pull request
Nov 15, 2026
* docs(work): spec for publishing research/ as a Pages site (#576) Starlight over a Python prep script so research/ stays the only copy of the notes; MkDocs Material is past its maintenance window and Zensical's hooks are partial. The local gate is left unchanged so research-only changes never need Bun. Fresh-review: two subagent passes over spec.md drafts, checked against NOTICE.md, gate.py, ADR 0038, renovate.jsonc, ci.yml, the python doc-tooling modules and research/ link and figure counts Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs(work): implementation plan for the research site (#576) Six tasks run as subagent dispatches: Sonnet for the Python prep step, the Starlight build and the prose pages; Haiku for the browser look and the Pages workflow, which the plan nearly transcribes; one Opus review of the finished branch, per docs/harness.md. configure-pages is left out because Astro's own Pages workflow doesn't need it and it would need a pages permission in the build job. Fresh-review: a Sonnet subagent read plan.md against spec.md, AGENTS.md, docs/harness.md § Policy, markdown_links.py, frontmatter.py, knowledge_map.py, gate.py, mise.toml, ci.yml, renovate.jsonc, NOTICE.md and docs/SPEC.md § 2.4, and recounted the notes, math, Mermaid and path figures under research/ Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * feat(site): page URLs and link rewriting for the research site (#576) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * feat(site): title, description and source link for each site page (#576) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * feat(site): prepare the research site's content from site/published.toml (#576) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * feat(site): Starlight build of the research site (#576) Bun 1.4.4 is pinned in mise.toml and mise.lock; site/ is a Bun project with exact versions. `mise run site` builds site/dist, `mise run site-serve` previews it. Differences from the plan's config, from the installed packages: - Astro 7 uses the Satteri Markdown processor by default; the remark/rehype plugins need @astrojs/markdown-remark (added, exact). Astro still accepts markdown.remarkPlugins/rehypePlugins but warns they are deprecated in favour of unified({...}); kept as the plan wrote them. - Starlight 0.43 removed a sidebar group's top-level `autogenerate`; the prep step now writes {label, items: [{autogenerate}]} (test updated). - starlight-links-validator drops only a lower-case `index`, so INDEX.md was unreachable at its folder URL; the prep step lower-cases written file names (test added). URLs are unchanged. - astro-mermaid 2.2.0 and starlight-links-validator 0.27.0 take the plan's usage as is (mermaid() before starlight, validator in `plugins`); the social shape and the `page.editLink` i18n key match too. - test_markdown_links_resolve skips the gitignored generated pages (site/src/content/docs/), whose site-absolute URLs the validator checks. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * fix(site): keep a leading code span in page titles; label hand-written pages by title (#576) page_text matched the heading on the code-blanked line, where the blanked span is absorbed by the whitespace after "#", so "# `PLAYONE.EXE`: X" became ": X" (22 pages). The title now comes from the original line. A page whose front matter sets title: also uses it as its sidebar label instead of the file stem. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * ci(site): build the research site on PRs and deploy it to Pages from main (#576) Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com> * docs(site): landing page and reading guide; comparison-only screenshots in NOTICE (#576) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * fix(site): normalise display math and fail the build on a KaTeX error (#576) micromark reads the remainder of a `$$` line as fence meta, so a multi-line block opened with `$$M = ...` swallowed the rest of coordinates-and-units and one-line `$$x$$` rendered inline. The prep step now puts each fence outside code on its own line, and a rehype plugin after rehype-katex throws on any message it recorded, so a bad formula cannot ship as a red span. Also: the unified() markdown processor replaces the deprecated plugin keys; a repo-root link opens the tree; the CLI and prepare name the file for IO and TOML errors; two sources writing one path (case-insensitively) raise; backslashes in published.toml paths are refused; a file section's label is its page title; descriptions lose Markdown; generated doc dirs get their own constant in the link test. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * docs(site): widen the licence and tag wording to match their sources (#576) reading.md now excepts cited community research and captured game frames from CC BY 4.0, as research/README.md, README.md and NOTICE.md do, gives [Superseded] a line and words the tags as SPEC section 2.4 does. NOTICE.md's comparison-sheet bullet covers the other layouts in work/. research/README.md tells note authors the notes are published and what the build needs. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * chore(site): group the site's npm packages in Renovate and bound the Pages jobs (#576) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> * docs(research): say when a note's file name breaks the site build (#576) page_url only checks names it is asked to link, so an unlinked note with a bad name still builds; the line now says the build fails when a published page links to it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs(work): link PR #578; Pages goes on before the first merge (#576) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(site): pin katex to the 0.16 rehype-katex renders with (#576) rehype-katex 7 renders with its own katex 0.16.47, which emits .sizing classes; the 0.19.1 stylesheet the site loaded styles .katex-sizing, so script-size math lost its sizing. Renovate now keeps katex below 0.17. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * fix(site): force one KaTeX 0.18.11 through a Bun override (#576) rehype-katex 7 still declares katex ^0.16.0 while mermaid takes ^0.18.0, so two copies coexisted and the 0.16.47 pin left the renderer on a release without the advisory fixes. The override gives the tree one KaTeX, 0.18.11, inside mermaid's range; the direct pin keeps the stylesheet the same version. Renovate now allows below 0.19. Drop the override once remarkjs/remark-math#122 ships. The math spike (work/2026-11/576_research-site/math-spike.md on claude/spike-site-math-b43fe3) has the evidence. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Initial checklist
Description of changes
Updates
katexinrehype-katexfrom^0.16.0to^0.19.0.katexbelow0.18.2is affected by GHSA-238p-pmpm-9mq7, and the current range keeps dependents from getting the fix without an override.Closes GH-121.
@types/katexis removed:katexships its own typeskatex@0.19.0: KaTeX0.18prefixed a number of internal class names (base→katex-baseand so on), so markup from this version needs the matching stylesheet\begin{split}\end{{split}}, which KaTeX now reports as aParseError(fix(environment): report invalid environment names as ParseError KaTeX/KaTeX#4256), so it no longer reached the fallback branch; it now uses\begin{array}{{c}}…, which still throws a plainErroron both0.16and0.19The
lib/code needs no change: it only usesrenderToString, and the options it passes (throwOnError,strict: 'ignore') behave the same.Because of the class name change I think this is a breaking change for users who load KaTeX’s CSS themselves, so it probably warrants a major. This matches the review on micromark/micromark-extension-math#14, which does the same update for the package
remark-mathuses.Note: locally (Node 24) the “should handle errors, support
errorColor” test fails on aVFileMessageshape difference; it fails the same way onmainwithout this change, so it is unrelated.